期刊
JOURNAL OF MEDICINAL CHEMISTRY
卷 65, 期 11, 页码 7918-7932出版社
AMER CHEMICAL SOC
DOI: 10.1021/acs.jmedchem.2c00460
关键词
-
资金
- National Key Research and Development Program of China [2021YFF1201400]
- Natural Science Foundation of Zhejiang Province [LZ19H300001, LD22H300001]
- Fundamental Research Funds for the Central Universities [2020QNA7003]
- Key R&D Program of Zhejiang Province [2020C03010]
The development of accurate machine-learning-based scoring functions for virtual screening requires unbiased and diverse datasets. However, most existing datasets may suffer from hidden biases and data insufficiency. In this study, we developed a new approach named TocoDecoy to generate unbiased and expandable datasets, and evaluated its performance compared to other datasets.
Development of accurate machine-learning-based scoring functions (MLSFs) for structure-based virtual screening against a given target requires a large unbiased dataset with structurally diverse actives and decoys. However, most datasets for the development of MLSFs were designed for traditional SFs and may suffer from hidden biases and data insufficiency. Hereby, we developed a new approach named Topology-based and Conformation-based decoys generation (TocoDecoy), which integrates two strategies to generate decoys by tweaking the actives for a specific target, to generate unbiased and expandable datasets for training and benchmarking MLSFs. For hidden bias evaluation, the performance of InteractionGraphNet (IGN) trained on the TocoDecoy, LIT-PCBA, and DUD-E-like datasets was assessed. The results illustrate that the IGN model trained on the TocoDecoy dataset is competitive with that trained on the LIT-PCBA dataset but remarkably outperforms that trained on the DUDE dataset, suggesting that the decoys in TocoDecoy are unbiased for training and benchmarking MLSFs.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据